Abstract
Quantitative positron emission tomography (PET) using statistical techniques requires: (a) a system geometric model that represents the probability of detecting an emission from each image pixel at each detector-pair, and (b) an iterative algorithm that reconstructs image as quantitative measurements of radiotracer distribution in vivo. Conventional implementations of iterative reconstruction use system geometric models based either on linear interpolation or on computing the volume of intersection of detection tubes with each voxel, but these simple models ignore many important physical system factors, like depth dependent geometric sensitivity and spatially variant detector pair resolution. In this paper, we evaluate a more accurate system geometric model that includes these physical factors. In addition, implementation variation among different iterative algorithms is another factor that limits the performance. Here, we compare performance of filtered backprojection (FBP) with the ordered subsets expectation maximization (OSEM) algorithm and a maximum a posteriori (MAP) method using a Gibbs prior with convex potential functions. Using the contrast recovery coefficient (CRC) as a performance measurement, we conducted various phantom experiments to investigate how the choices of algorithm and system matrix affect reconstruction accuracy. The results of these studies show that all of the iterative methods tested produce superior CRCs than FBP at matched background variance. And the combination of the accurate system geometric model and MAP reconstruction algorithm outperforms the other statistical methods.